Storing and Using Past Plans and Negotiation Results in Multiagent Systems

T. Sugawara, O. Akashi, and S. Kurihara (Japan)

Keywords

Distributed artificial intelligence,planning, plan reuse, casebased reasoning,coordination, conflict resolution

Abstract

This paper describes plan reuse in multiagent domains where each agent has its own goals to achieve, and its tasks may affect or be affected by other agents working simultaneously. The basic idea of plan reuse, where past planning results are reused for a new problem, was shown in [1]. This paper extends that work to exploit the nature of hierarchical planning. First we show that a hierarchical structure of planning can make it easier to verify the similarity of problems and reduce the cost of retrieving plans. Then we add annotations at a higher level of the hierarchy in order to detect conflicts at an earlier stage. This can eliminate the backtracking that may occur in planning with plan templates. Finally, this paper shows, through examples, how annotations for higher levels are extracted and work to avoid some conflicts.

Important Links:



Go Back